snowflake-mcp
Enables AI agents to execute SQL queries and explore Snowflake databases using natural language, with schema discovery, table inspection, and readonly mode.
README
<p align="center"> <img src="assets/logo-600.png" alt="Snowflake MCP" width="600"> </p>
Snowflake MCP Server
A Model Context Protocol (MCP) server that enables AI agents to execute SQL queries against Snowflake databases.
Users can use natural language to query Snowflake databases, like:
- "Get me the top 10 products by revenue"
- "Show me the total revenue for the last 30 days"
- "Describe the structure of the orders table"
- "Explore the database and summarize what data is available"
- "Build a query to find customers who haven't ordered in 90 days"
Features
- Execute SQL queries directly from AI agents like Cursor, Claude Desktop, etc.
- Schema Discovery: Browse databases, schemas, tables, and views
- Table Inspection: Describe table structures, view sample data, check row counts
- Query Safety: Readonly mode, row limits, and query timeouts
- Multiple Output Formats: Table, JSON, or CSV
- Query Explanation: Get execution plans for queries
- MCP Prompts: Guided workflows for common tasks
- Support for both password and SSO (browser-based) authentication
- Configurable default warehouse, database, schema, and role
Prerequisites
- Node.js 18 or later
Usage
Cursor
Add to your Cursor MCP settings (~/.cursor/mcp.json):
SSO Authentication (Recommended)
{
"mcpServers": {
"snowflake": {
"command": "npx",
"args": ["-y", "snowflake-mcp"],
"env": {
"SNOWFLAKE_ACCOUNT": "your-org-your-account",
"SNOWFLAKE_USERNAME": "your-username",
"SNOWFLAKE_AUTHENTICATOR": "externalbrowser",
"SNOWFLAKE_WAREHOUSE": "your-warehouse",
"SNOWFLAKE_DATABASE": "your-database",
"SNOWFLAKE_SCHEMA": "your-schema"
}
}
}
}
A browser window will open for authentication on first query.
Password Authentication
{
"mcpServers": {
"snowflake": {
"command": "npx",
"args": ["-y", "snowflake-mcp"],
"env": {
"SNOWFLAKE_ACCOUNT": "your-org-your-account",
"SNOWFLAKE_USERNAME": "your-username",
"SNOWFLAKE_PASSWORD": "your-password",
"SNOWFLAKE_ROLE": "your-role",
"SNOWFLAKE_WAREHOUSE": "your-warehouse",
"SNOWFLAKE_DATABASE": "your-database",
"SNOWFLAKE_SCHEMA": "your-schema",
"SNOWFLAKE_READONLY": "true"
}
}
}
}
Claude Desktop
Add to your Claude Desktop config (~/Library/Application Support/Claude/claude_desktop_config.json on macOS):
{
"mcpServers": {
"snowflake": {
"command": "npx",
"args": ["-y", "snowflake-mcp"],
"env": {
"SNOWFLAKE_ACCOUNT": "your-org-your-account",
"SNOWFLAKE_USERNAME": "your-username",
"SNOWFLAKE_AUTHENTICATOR": "externalbrowser"
}
}
}
}
Configuration
| Variable | Required | Default | Description |
|---|---|---|---|
SNOWFLAKE_ACCOUNT |
Yes | - | Your Snowflake account identifier (e.g., ORG-ACCOUNT) |
SNOWFLAKE_USERNAME |
Yes | - | Snowflake username |
SNOWFLAKE_AUTHENTICATOR |
No | externalbrowser |
Authentication method: externalbrowser (SSO) or snowflake (password) |
SNOWFLAKE_PASSWORD |
Conditional | - | Required if authenticator is snowflake, not needed for externalbrowser |
SNOWFLAKE_ROLE |
No | - | Role to use for the session (uses account default if not set) |
SNOWFLAKE_WAREHOUSE |
No | - | Warehouse to use (uses account default if not set) |
SNOWFLAKE_DATABASE |
No | - | Database to use (uses account default if not set) |
SNOWFLAKE_SCHEMA |
No | - | Schema to use (uses account default if not set) |
SNOWFLAKE_READONLY |
No | false |
Set to true to block write operations (INSERT, UPDATE, DELETE, etc.) |
Finding Your Connection Settings
You can find your connection settings in Snowsight (Snowflake's web interface):
- Sign in to Snowsight
- Click on your username in the bottom-left corner to open the user menu
- Select Connect a tool to Snowflake
- Open the
Config Filetab - Select the Warehouse, Database, Schema you want to use
- Copy values from the generated config file
Available Tools
Query Execution
execute_query
Execute a SQL query against Snowflake.
| Parameter | Type | Required | Default | Description |
|---|---|---|---|---|
query |
string | Yes | - | The SQL query to execute |
max_rows |
number | No | 100 | Maximum number of rows to return |
timeout |
number | No | - | Query timeout in seconds |
format |
string | No | table | Output format: table, json, or csv |
explain_query
Get the execution plan for a SQL query without running it.
| Parameter | Type | Required | Description |
|---|---|---|---|
query |
string | Yes | The SQL query to explain |
Connection
test_connection
Test the connection to Snowflake and return connection info including current user, role, warehouse, database, schema, and version.
Schema Discovery
list_databases
List all accessible databases in Snowflake.
list_schemas
List all schemas in a database.
| Parameter | Type | Required | Description |
|---|---|---|---|
database |
string | No | Database name (uses current if not specified) |
list_tables
List all tables in a schema.
| Parameter | Type | Required | Description |
|---|---|---|---|
database |
string | No | Database name |
schema |
string | No | Schema name |
list_views
List all views in a schema.
| Parameter | Type | Required | Description |
|---|---|---|---|
database |
string | No | Database name |
schema |
string | No | Schema name |
Table Inspection
describe_table
Get detailed information about a table's structure including columns, types, and constraints.
| Parameter | Type | Required | Description |
|---|---|---|---|
table |
string | Yes | Table name (can include database.schema.table) |
get_table_sample
Get a sample of rows from a table to understand its data.
| Parameter | Type | Required | Default | Description |
|---|---|---|---|---|
table |
string | Yes | - | Table name |
limit |
number | No | 5 | Number of sample rows to return |
get_table_row_count
Get the total number of rows in a table.
| Parameter | Type | Required | Description |
|---|---|---|---|
table |
string | Yes | Table name |
get_primary_keys
Get primary key columns for a table.
| Parameter | Type | Required | Description |
|---|---|---|---|
table |
string | Yes | Table name |
MCP Resources
schema://current
Returns information about the current database schema, including:
- Current database and schema names
- List of all tables
- List of all views
Access this resource to get a quick overview of the connected schema without running queries.
MCP Prompts
analyze_table
Analyze a table's structure, sample data, and get query suggestions.
| Parameter | Type | Required | Description |
|---|---|---|---|
table_name |
string | Yes | The table to analyze |
explore_database
Explore and summarize the structure of a database.
| Parameter | Type | Required | Description |
|---|---|---|---|
database_name |
string | No | Database to explore (uses current if not specified) |
query_builder
Help build a SQL query based on natural language description.
| Parameter | Type | Required | Description |
|---|---|---|---|
description |
string | Yes | Natural language description of desired query |
tables |
string | No | Comma-separated list of relevant tables |
Security Considerations
- SSO authentication is recommended for production use as it avoids storing passwords in configuration files
- Readonly mode (
SNOWFLAKE_READONLY=true) is recommended when you only need to query data - Never commit configuration files containing credentials to version control
- Consider using environment variables or a secrets manager for sensitive values
- The server executes queries with the permissions of the configured Snowflake user—ensure appropriate access controls are in place
- The
max_rowsparameter helps prevent accidentally returning massive result sets
License
MIT License
推荐服务器
Baidu Map
百度地图核心API现已全面兼容MCP协议,是国内首家兼容MCP协议的地图服务商。
Playwright MCP Server
一个模型上下文协议服务器,它使大型语言模型能够通过结构化的可访问性快照与网页进行交互,而无需视觉模型或屏幕截图。
Magic Component Platform (MCP)
一个由人工智能驱动的工具,可以从自然语言描述生成现代化的用户界面组件,并与流行的集成开发环境(IDE)集成,从而简化用户界面开发流程。
Audiense Insights MCP Server
通过模型上下文协议启用与 Audiense Insights 账户的交互,从而促进营销洞察和受众数据的提取和分析,包括人口统计信息、行为和影响者互动。
VeyraX
一个单一的 MCP 工具,连接你所有喜爱的工具:Gmail、日历以及其他 40 多个工具。
graphlit-mcp-server
模型上下文协议 (MCP) 服务器实现了 MCP 客户端与 Graphlit 服务之间的集成。 除了网络爬取之外,还可以将任何内容(从 Slack 到 Gmail 再到播客订阅源)导入到 Graphlit 项目中,然后从 MCP 客户端检索相关内容。
Kagi MCP Server
一个 MCP 服务器,集成了 Kagi 搜索功能和 Claude AI,使 Claude 能够在回答需要最新信息的问题时执行实时网络搜索。
e2b-mcp-server
使用 MCP 通过 e2b 运行代码。
Neon MCP Server
用于与 Neon 管理 API 和数据库交互的 MCP 服务器
Exa MCP Server
模型上下文协议(MCP)服务器允许像 Claude 这样的 AI 助手使用 Exa AI 搜索 API 进行网络搜索。这种设置允许 AI 模型以安全和受控的方式获取实时的网络信息。